[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-81323":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":23,"hasPages":23,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":34,"readmeContent":35,"aiSummary":36,"trendingCount":16,"starSnapshotCount":16,"syncStatus":15,"lastSyncTime":37,"discoverSource":38},81323,"loopat","simpx\u002Floopat","simpx","Self-hosted AI workspace built around context management. Works solo, scales to teams.","",null,"TypeScript",57,11,44,2,0,3,6,13,9,53.54,"Apache License 2.0",false,"main",[26,27,28,29,30,31,32,33],"agentic-ai","ai-agent","claude","claude-agetn-sdk","coding-agent","context-management","developer-tools","team-collaboration","2026-06-12 04:01:32","# 🧶 loopat\n\n> **Self-hosted AI workspace built around context management — works solo, scales to teams**\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Fscreenshot.png\" alt=\"loopat — Loop view with chat, workdir, terminal, and team DM\" width=\"100%\">\n\u003C\u002Fp>\n\n⭐ [Star on GitHub](https:\u002F\u002Fgithub.com\u002Fsimpx\u002Floopat) ·\n🚀 [Quick start](#quick-start) ·\n📖 [Architecture](docs\u002Farchitecture.md)\n\n---\n\nWhen humans collaborate with AI, three things only humans can bring:\n\n- **Drive** — pushing the work forward. AI has no desires, no\n  ambition of its own; momentum has to come from a human.\n- **Attention** — what matters now, what to ignore. AI doesn't know\n  what's worth your time.\n- **Entropy reduction** — turning noise into structured knowledge.\n  AI generates tokens but won't spontaneously simplify.\n\nloopat is built around managing these three as first-class concepts:\n**Loop** (drive) · **Focus** (attention) · **Context** (entropy\nreduction). A fourth concept — **Chat** — coordinates the team on the\nsync axis.\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"docs\u002Foverview.svg\" alt=\"loopat architecture\" width=\"100%\">\n\u003C\u002Fp>\n\nThe agent itself is the [Claude Agent SDK][sdk]; what makes loopat\ndistinct is the **context architecture around it** — how chat, code,\nmemory, and knowledge interlock so context doesn't get lost across\nsessions or teammates.\n\n[sdk]: https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-agent-sdk\n\n---\n\n## What makes loopat different\n\n- **End-to-end context management.** Team chat (IM) threads, code\n  edits, agent decisions, memory — all live in the same context graph\n  and all flow into the next loop. Most AI tools make you copy-paste\n  from Slack into the AI to give it situational context; loopat treats\n  team chat as **a first-class context source** — spawn a loop from\n  any chat thread and that thread becomes part of the loop's context\n  automatically.\n- **Works solo, scales to teams.** Same workspace whether you're\n  alone or onboarding teammates. Solo, it's a personal AI workspace;\n  with a team, shared `knowledge\u002F` and `notes\u002F` git repos sync across\n  members, loops auto-commit their work, and observations promote\n  upward through a distillation pipeline. Most AI tools force you to\n  pick between solo CLI and team SaaS — loopat is one tool at any scale.\n- **Reproducible loops.** Every loop runs in its own sandbox with a\n  versioned toolchain and a pinned credential vault. Spawn the same\n  loop tomorrow on a different machine and get the same starting\n  state. No \"works on my machine\" for AI sessions.\n- **Self-hosted, data you own.** All artifacts live in plain git\n  repos you fully control; vault secrets are git-crypt encrypted.\n  BYO API key — nothing leaves your machine except the model API call\n  itself.\n\n## How loopat compares\n\n| | Claude Code | Cursor | opencode | Codex | **loopat** |\n|---|---|---|---|---|---|\n| Form factor | CLI | IDE | TUI | Web (hosted) | **Web (self-hosted)** |\n| License | proprietary | proprietary | MIT | proprietary | **Apache 2.0** |\n| Team IM integration | external (manual paste) | external (manual paste) | external (manual paste) | external (manual paste) | **built-in, threads ingested into loop context** |\n| Memory management | personal (`CLAUDE.md`) | personal (rules + memories) | personal (`AGENTS.md`) | none | **personal + team-shared, with distillation pipeline** |\n| Multi-user | single user | per-seat | single user | per-account | **shared workspace** |\n| Shared team knowledge | individual config | individual config | individual config | individual config | **git-synced across team** |\n| Per-session sandbox | process-level | process-level | process-level | OpenAI-managed | **bwrap (default) · Docker (planned)** |\n| Toolchain pinning | host runtime | host runtime | host runtime | fixed (hosted env) | **per-loop versioned** |\n| Per-task credential isolation | shared (env vars) | shared (subscription) | shared (env vars) | account-managed | **per-loop vault overlay** |\n| Data location | local files | cloud | local files | OpenAI servers | **git repos you control** |\n| Secrets storage | env vars (plaintext) | cloud-managed | env vars (plaintext) | platform-managed | **git-crypt encrypted vault** |\n| Agent engine | proprietary (Anthropic) | proprietary (multi-model) | pluggable | proprietary (OpenAI) | **Claude Agent SDK** |\n\n---\n\n## Quick start\n\n```sh\ngit clone https:\u002F\u002Fgithub.com\u002Fsimpx\u002Floopat.git\ncd loopat && bun install\nbun run dev\n```\n\nOpen \u003Chttp:\u002F\u002Flocalhost:7787>. The first run bootstraps `~\u002F.loopat\u002F`,\nprints a checklist, and prompts you to set your API key in\n`~\u002F.loopat\u002Fconfig.json`. Restart — done.\n\n> **Needs:** Linux + [bubblewrap][bwrap] + [mise][mise] + [bun][bun] on\n> the host. macOS \u002F Windows is via Docker (see below). For team setups\n> with shared knowledge\u002Fnotes git repos and full bootstrap details, see\n> the [installation guide](docs\u002Finstall.md).\n\n### Setup guides\n\nLoopat splits configuration along role lines — read whichever applies:\n\n- **[Admin setup](docs\u002Fsetup-admin.md)** — provision the workspace:\n  knowledge \u002F notes repos, team sandboxes, MCP, operator mounts. Run\n  this once per workspace.\n- **[User setup](docs\u002Fsetup-user.md)** — join an existing workspace:\n  personal credential repo, providers, vaults, sandbox mounts. Run\n  this once per member, per machine.\n\n[bwrap]: https:\u002F\u002Fgithub.com\u002Fcontainers\u002Fbubblewrap\n[mise]: https:\u002F\u002Fmise.jdx.dev\u002F\n[bun]: https:\u002F\u002Fbun.sh\u002F\n\n## Deployment\n\n### Docker (recommended)\n\n```sh\ndocker compose up -d\n```\n\nOpen \u003Chttp:\u002F\u002Flocalhost:17787> (note: **17787**, not 7787 — the host\nport is remapped to avoid collision with a local dev server; see\n[`docker-compose.yml`](docker-compose.yml)). Workspace persists in\nthe `loopat-data` volume. Needs `SYS_ADMIN` + unconfined AppArmor for\nbwrap mount namespaces.\n\n### From source (Linux)\n\n```sh\nbun run build                     # installs deps + builds frontend → web\u002Fdist\u002F\nPORT=7787 bun run server\u002Fsrc\u002Findex.ts\n```\n\nSingle Hono process serves API + static SPA + websocket on one port.\nPut a reverse proxy in front and proxy `\u002Fapi` + `\u002Fws` to the server.\n\n## Documentation\n\n- **[Admin setup](docs\u002Fsetup-admin.md)** — workspace config, knowledge \u002F\n  notes repos, sandboxes, MCP, operator mounts.\n- **[User setup](docs\u002Fsetup-user.md)** — personal repo, providers,\n  vaults, sandbox mounts.\n- **[Installation guide](docs\u002Finstall.md)** — host install, system deps,\n  environment variables.\n- **[Architecture](docs\u002Farchitecture.md)** — the read\u002Fwrite path, layered\n  context model, distillation pipeline, Claude config injection paths.\n- **[Context flow](docs\u002Fcontext-flow.md)** — the horizontal working model: a\n  loop is a git worktree, shared context is `main`, and loops exchange it over\n  two edges — pull and promote.\n- **[Identity](docs\u002Fidentity.md)** — who a loop acts as: the credential chain\n  (deploy key → git-crypt → vault), and how loopat integrates with your git\n  host so onboarding is one authorization.\n- **[.claude composition](docs\u002Fcomposition.md)** — how team \u002F profile \u002F\n  personal \u002F repo `.claude\u002F` tiers merge into the loop runtime, and what\n  you can put in each tier.\n- **[Sandbox](docs\u002Fsandbox.md)** — bwrap mount mechanics, three-tier mount\n  authority, what stops the agent from escaping.\n- **[Troubleshooting](docs\u002Ftroubleshoot.md)** — chat won't start, banner\n  errors, common pitfalls.\n\n## Contributing\n\nIssues and PRs welcome. Before opening a non-trivial PR, please skim\n[`docs\u002Farchitecture.md`](docs\u002Farchitecture.md) so the change lands in the\nright layer (sandbox \u002F vault \u002F loop \u002F chat).\n\nContributors are asked to sign the [Contributor License Agreement](CLA.md)\non their first PR — the [CLA Assistant][cla-assistant] bot prompts you\nwith a one-click link. This keeps future licensing options open (e.g.\nmoving to a business-friendly license) without having to re-collect\npermission from every contributor.\n\n[cla-assistant]: https:\u002F\u002Fcla-assistant.io\u002F\n\n## Acknowledgments\n\nloopat is built on top of:\n\n- [Claude Agent SDK][sdk] — the agent runtime\n- [assistant-ui](https:\u002F\u002Fgithub.com\u002Fassistant-ui\u002Fassistant-ui) — React\n  components for the chat interface\n- [Hono](https:\u002F\u002Fhono.dev\u002F) — the HTTP + WebSocket server\n- [bubblewrap](https:\u002F\u002Fgithub.com\u002Fcontainers\u002Fbubblewrap) — sandbox mount\n  namespaces\n- [mise](https:\u002F\u002Fmise.jdx.dev\u002F) — per-loop toolchain installs\n\n## License\n\n[Apache License 2.0](LICENSE). See [`NOTICE`](NOTICE) for required\nattributions and [`CLA.md`](CLA.md) for contribution terms.\n","loopat 是一个围绕上下文管理构建的自托管AI工作空间，既适合个人使用也能扩展至团队协作。该项目的核心功能包括端到端的上下文管理、可重现的工作循环以及对团队即时消息集成的支持，使得从聊天线程到代码编辑的所有信息都能无缝地融入工作流程中，确保跨会话和成员间的信息连贯性。采用TypeScript编写，并基于Claude Agent SDK，通过版本控制工具链和加密凭证库来保证工作的可复现性和数据安全性。适用于需要高效整合人类与AI合作成果的开发场景，特别是在注重隐私保护和自主控制数据的企业环境中表现尤为出色。","2026-06-11 04:04:36","CREATED_QUERY"]